Multi-head co-training: An uncertainty-aware and robust semi-supervised learning framework

计算机科学 机器学习 人工智能 一般化 集合(抽象数据类型) 架空(工程) 渲染(计算机图形) 数据挖掘 数学 数学分析 程序设计语言 操作系统
作者
Mingcai Chen,Chongjun Wang
出处
期刊:Knowledge Based Systems [Elsevier]
卷期号:302: 112325-112325
标识
DOI:10.1016/j.knosys.2024.112325
摘要

Co-training, an advanced form of self-training, allows multiple base models to learn collaboratively, leading to superior performance in semi-supervised learning tasks. However, its widespread adoption is hindered by high computational costs and intricate design choices. To address these challenges, we present Multi-Head Co-Training, a streamlined and efficient framework that consolidates individual models into a multi-head structure, adding minimal extra parameters. Each classification head in this unified model collaborates with others via a "Weak and Strong Augmentation" strategy, with diversity organically introduced through robust data augmentation. Consequently, our approach implicitly promotes diversity while incurring only a minor increase in computational overhead, making co-training more accessible. We validate the effectiveness of Multi-Head Co-Training through an empirical study on standard semi-supervised learning benchmarks. For example, our method achieves up to a 3.1% accuracy improvement on the semi-supervised CIFAR dataset compared to recent methods. Recognizing the necessity for more practical performance metrics beyond accuracy, we assess our framework from three additional perspectives: robust generalization, uncertainty, and computational efficiency. To evaluate robust generalization, we expand the conventional SSL experimental setting to a more comprehensive open-set semi-supervised learning scenario. For uncertainty assessment, we conduct experiments on model calibration and selective classification benchmarks. For example, our method achieves up to a 4.3% accuracy improvement on the open-set semi-supervised CIFAR dataset. Our extensive experiments confirm that our proposed framework better captures prediction confidence and uncertainty, rendering it more suitable for SSL deployment in open environments. The code is available at https://github.com/chenmc1996/Multi-Head-Co-Training.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
更新
大幅提高文件上传限制,最高150M (2024-4-1)

科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
西红柿炒番茄应助honghong采纳,获得30
刚刚
刚刚
Mars夜愿发布了新的文献求助10
1秒前
1秒前
陈陈快点去读书完成签到,获得积分10
1秒前
五六七发布了新的文献求助10
2秒前
蜻蜓发布了新的文献求助10
2秒前
小蘑菇应助PROPELLER采纳,获得10
2秒前
小蘑菇应助高岩采纳,获得10
3秒前
打打发布了新的文献求助10
3秒前
YJ发布了新的文献求助10
4秒前
张张发布了新的文献求助10
5秒前
qly发布了新的文献求助10
6秒前
Mars夜愿完成签到,获得积分10
7秒前
tuanhust完成签到,获得积分10
7秒前
taster发布了新的文献求助10
9秒前
英姑应助小蚊子采纳,获得10
9秒前
10秒前
谢文强完成签到,获得积分10
10秒前
oywc应助刘斌采纳,获得10
11秒前
11秒前
12秒前
12秒前
慕青应助青枫也燃烧采纳,获得10
13秒前
White Night完成签到,获得积分10
14秒前
欣喜的听枫完成签到,获得积分10
14秒前
Zz完成签到 ,获得积分0
14秒前
完美世界应助张张采纳,获得10
15秒前
科研通AI2S应助科研小白采纳,获得10
15秒前
贝湾发布了新的文献求助10
16秒前
16秒前
Debjor发布了新的文献求助10
16秒前
科研通AI2S应助马迦南采纳,获得10
16秒前
taster完成签到,获得积分10
17秒前
fifteen发布了新的文献求助10
17秒前
17秒前
无名完成签到,获得积分10
18秒前
易小名完成签到,获得积分10
18秒前
18秒前
赘婿应助shai_ga采纳,获得10
18秒前
高分求助中
Evolution 10000
Sustainability in Tides Chemistry 2800
юрские динозавры восточного забайкалья 800
English Wealden Fossils 700
An Introduction to Geographical and Urban Economics: A Spiky World Book by Charles van Marrewijk, Harry Garretsen, and Steven Brakman 600
Diagnostic immunohistochemistry : theranostic and genomic applications 6th Edition 500
Mantiden: Faszinierende Lauerjäger Faszinierende Lauerjäger 400
热门求助领域 (近24小时)
化学 医学 生物 材料科学 工程类 有机化学 生物化学 物理 内科学 纳米技术 计算机科学 化学工程 复合材料 基因 遗传学 催化作用 物理化学 免疫学 量子力学 细胞生物学
热门帖子
关注 科研通微信公众号,转发送积分 3154185
求助须知:如何正确求助?哪些是违规求助? 2805059
关于积分的说明 7863283
捐赠科研通 2463232
什么是DOI,文献DOI怎么找? 1311173
科研通“疑难数据库(出版商)”最低求助积分说明 629464
版权声明 601821